trellis-continue

Resumes in-progress tasks by loading the workflow phase index and routing to the correct step.

Updated Jun 8, 2026
One-click install
npx skills add https://github.com/dev2019zheng/codex-beacon --skill trellis-continue-dev2019zheng
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: trellis-continue
Source: https://github.com/dev2019zheng/codex-beacon/tree/main/.agents/skills/trellis-continue
Command: npx skills add https://github.com/dev2019zheng/codex-beacon --skill trellis-continue-dev2019zheng

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When returning to an in-progress task, it is easy to lose track of which phase and step you were on. This Skill restores that context by reading the task status and artifacts, then routing you to the exact workflow step to continue from. ## Core Features & Use Cases - Context Restoration: Runs get_context.py to confirm the current task, git state, and recent commits before resuming work. - Status-Based Routing: Maps task status (planning, in_progress, completed) plus artifact presence (prd.md, implement.jsonl) to the correct phase step such as 1.1, 1.3, 2.1, or 3.1. - Step-Level Instructions: Loads detailed guidance for a specific step via get_context.py --mode phase --step, so you follow the canonical workflow rather than guessing. - Use Case: You return to a feature task after a day away. Run this Skill to discover the task is in planning with a finished prd.md but an uncurated implement.jsonl, so you resume directly at step 1.3. ## Quick Start Ask the assistant to continue the current Trellis task and load the correct workflow step to resume from.

Frequently Asked Questions about trellis-continue

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I resume an in-progress task after a break?

Run get_context.py to confirm the current task and git state, then load the phase index with --mode phase. The Skill routes you to the correct step based on the task status field and which artifacts already exist.

How does the skill decide which workflow step to continue at?

It reads the task status field and checks artifact presence. For example, status=planning with an existing prd.md but an uncurated implement.jsonl routes to step 1.3, while status=in_progress with a passed check routes to step 3.1.

Can I skip workflow steps that are already done?

Yes. Steps marked [once] are skipped if their output already exists, such as prd.md for step 1.1 or a curated implement.jsonl for step 1.3. Steps marked [required] must still run in order within a phase.

What happens if I need to go back to an earlier phase?

The workflow allows returning to an earlier phase when discoveries require it. The full phase rules, routing table, and do-not-skip list live in .trellis/workflow.md, which is the canonical reference.

Does this skill work without the Trellis workflow files?

No. It depends on .trellis/scripts/get_context.py and .trellis/workflow.md to determine task state and load step instructions. Without those files, there is no phase index or status to route from.